TressFX 2.0 : Improved HAIR PHYSICS , Less PERFORMANCE HIT..YEAH Even On GTX CARDStressfx 2.0 is here .
DEMO VIDEO : http://www.youtube.com/watch?v=BftE-xwooZk
DEMO download link : http://developer.amd.com/wordpress/m...sFX11_v2.0.zip (8 MB)
Qoute from PCGamesHardware.com
Souece : http://www.pcgameshardware.de/AMD-Ra...-Demo-1104056/
The realistic representation of hair , for example, in computer games is very difficult to achieve with conventional methods. In virtually all cases, the result is difficult to be desired , often see the usually formed from polygons and alpha textures and rather sparsely animated virtual hairstyles far from lifelike . Perhaps one reason why characters often wear a boring crewcut in games or have the same shine the bald head. AMD tried to tackle the problem with TressFX ( tress = Locke ), a technology that utilizes the computing power of the graphics card to represent thousands of individual strands and to encourage a physically correct . The technology is open , so it can also be illustrated with graphics cards from other manufacturers.
TressFX 2.0 - problems of the previous version
The technique was first in the reboot of Tomb Raider in spring 2013 - and so far only there - used to represent virtual Lara's hair. The result could be quite impressive , but some major problems prevented a storm of enthusiasm of the player base . Firstly, the effect required quite a lot of power , which is especially Nvidia users affected negatively because of GeForce GPUs broke the frame rates due to lower compute power much stronger than on a comparable Radeons . Furthermore, the hair behaved not physically credible in many situations. Although AMD improved by using a patch , but the basic problem remained. Thus the mane swirled frequently due to lack of ( virtual ) weight in unrealistic ways , the simulated elasticity of the hair caused that strands behaved with fast movements like rubber bands .
TressFX 2.0 - Improved Performance
With TressFX 2.0 , these problems are now a thing of the past. The performance issue has defused with new , finely resolved and yet more economical to be calculated shadows and a LoD system AMD. The shadows are now calculated only where they will be seen , the same applies to the hair strands . Furthermore, the removal of hair will now be taken into consideration for the camera. In the previous version , the hair was completely calculated , so also in the case when the figure was far away from the camera and limited by the resolution not all strands could be shown . This was fixed with TressFX 2.0. A classic LOD system as polygon models work by the hair not , therefore, another gimmick was used: Removes the camera from the head of hair , the number of hair strands is reduced , but their thickness increases at the same time . Thus, the illusion of a consistent detail level can be realized at the same time the computational cost but can be greatly reduced . According to AMD were more than halved with these adjustments, the performance penalty .
TressFX 2.0 - Easier to implement
In order to provide designers a greater incentive to install TressFX in their game, also the implementation was simplified. To minimize unrealistic turbulence and the rubber band effect , TressFX 2.0 can now more extensive and easier to adjust . The hair can be already configured in the tech demo and be set or changed such as the thickness , transparency, color and severity of the individual strands. If we precisely draw these possibilities even consider , for example, wet hair could be relatively easily simulated: The color slightly darker , redeem the transparency of something and clearly give the hair more weight - a simple " wet effect" would be complete without programming feasible. This of course is not the end , the programmer can of course also make deeper interventions .
TressFX 2.0 - The Demo
We have tried the demo on a GTX titanium, with standard settings requires the powerful Nvidia GPU in 2560 x 1440 just over two milliseconds for the calculation of the effect. Consider a frame rate of 60 frames per second without TressFX 2.0 out would equal the nearly eight frames that would be lost by switching the effect. To counter testing we planted following a Radeon R9 290 in our test system . Not surprisingly, the AMD GPU is significantly faster in the calculation , the performance loss by TressFX is compared to the GTX Titan by almost 25 % lower.
As superior in direct computing , R9 290 gets 25% less performance hit than GTX titan when tessfx is switched on.